CN106875326B - Method for hiding and extracting audio anti-counterfeiting signal in printed image - Google Patents

Abstract

The present invention aims at the problems of the current packaging and printing anti-counterfeiting technology that uses binary or pseudo-random scrambled sequences as watermark information, the information is single, and the robustness and invisibility need to be improved after printing-scanning. Combining audio information and packaging anti-counterfeiting, a method for hiding and extracting audio anti-counterfeiting signals in printed images is proposed. The anti-counterfeiting method for color printed products by leaf transformation includes: (I) making audio anti-counterfeiting information as watermark information, (II) confirming the watermark embedding position in the host image, (III) transforming to obtain a color image containing audio watermark information, and using it for Printing plate-making output; and extracting audio anti-counterfeiting information from printed images are two processes. This method combines the auditory characteristics of human ears and has a good balance in the combination of watermark robustness and invisibility.

Description

A method for hiding and extracting audio anti-counterfeiting signals in printed images

technical field

The invention relates to an audio anti-counterfeiting method, in particular to a method for hiding and extracting an audio anti-counterfeiting signal in a printed image, and belongs to the technical field of anti-counterfeiting of packaged printed matter.

Background technique

At present, counterfeit and shoddy goods have a negative impact on people's lives and economic development. Therefore, it is particularly important to use scientific and technological means to study anti-counterfeiting technology and prevent the harm caused by counterfeit goods. At present, the anti-counterfeiting packaging of commodities mostly relies on the anti-counterfeiting of packaging materials and anti-counterfeiting code design. Although these anti-counterfeiting technologies have been developed to a certain extent, counterfeiters can still copy and imitate some anti-counterfeiting technologies through various means. Many disadvantages.

The holographic watermark anti-counterfeiting technology has good robustness and security, and can be used as a green and intelligent anti-counterfeiting technology in the field of packaging anti-counterfeiting to reduce the cost of anti-counterfeiting packaging. Effective utilization of packaging. In order to realize the application of watermarking anti-counterfeiting methods in printed images, the performance of the watermarking algorithm, such as the robustness, is mostly optimized and innovated. Compared with the visual perception system under image information, the human ear can accept extremely dynamic sounds of different intensities. , through the small time difference (distance, time, frequency difference) between the direct sound and the reflected sound through the outer ear to distinguish the sound sources in different directions, and can also automatically select the sound source of interest in a complex noise environment.

Through the search of domestic patents, it is found that there are related patents, such as the following:

1. Patent application number CN201380010689.X, the invention patent titled "Audible Credential Recognition for Visually Impaired Persons" discloses a credential or article that carries the information for audible identification of the credential or article, wherein The information is located in or on the credential or item in the form of a spectral density function of frequency versus time (a spectrogram) implemented using credential anti-counterfeiting tools. Further disclosed is a method for making the voucher or item; a reading device for representing audible authentication information from the voucher or item, a method for authenticating the voucher or item, and a method for authenticating the voucher or item Use of spectrograms for identification purposes.

2. Patent application number CN200810196257.3, the invention patent titled "Method for Making and Recognition of Digital Watermarks for Audio Files" discloses a method for making and recognizing digital watermarks for audio files. A watermark is added as copyright marking information in the audio digital file to be processed, and it is characterized in that: the watermark is realized by using the converted voiceprint information, including: collecting a piece of voice signal identifying copyright and obtaining a voiceprint watermark image; converting it into a Two-dimensional binary sequence group, the gray value of each matrix unit is represented by an 8-bit binary number; the audio file to be inserted into the watermark is segmented, and each segment of audio frequency is transformed in the frequency domain to obtain a spectrum file; in each segment of the spectrum file Embed one-bit one-dimensional binary sequence group watermark data in it, and the embedding method is a method of quantifying and modifying the coefficients of the spectrum file; perform inverse transformation to obtain the audio data embedded in the watermark, and combine with the remaining unprocessed audio data to obtain a Digitally watermarked audio digital files. Reverse separation operation is performed during identification. The invention effectively solves the copyright authentication problem of digital audio products.

Although the above patents all involve adding audio as a watermark, they do not have the solution of hiding the audio as watermark information in the printing and packaging anti-counterfeiting image, and do not give full play to the advantages of audio reception by the human ear.

SUMMARY OF THE INVENTION

The invention aims at the problems of the current packaging and printing anti-counterfeiting technology that uses binary or pseudo-random scrambled sequences as watermark information, the information is single, and the robustness and invisibility need to be improved after printing-scanning. Combining audio information and packaging anti-counterfeiting, a method for hiding and extracting audio anti-counterfeiting signals in printed images is proposed. The anti-counterfeiting method of color printing products based on leaf transformation, the method combines the auditory characteristics of the human ear, and has a good balance effect in the combination of watermark robustness and invisibility.

The technical means adopted by the present invention to solve the above problems are: a method for hiding and extracting audio anti-counterfeiting signals in a printed image, including hiding the audio anti-counterfeiting signal in the printed image and extracting two audio anti-counterfeiting information from the printed image. The process; in which the audio anti-counterfeiting signal is hidden in the printed image, that is, the audio is used as the watermark anti-counterfeiting information, and the image is used as the host information, including: (I) making the audio anti-counterfeiting information as watermark information, (II) confirming the embedding of the watermark in the host image (III) Transform to obtain a color image containing audio watermark information, and use it for printing plate-making output; extracting audio anti-counterfeiting information from the printed image is the hidden inverse process, including: (A) Obtaining the image with an imaging device, (B) ) to perform information transformation after image correction, (C) to restore audio anti-counterfeiting information after holographic reproduction.

Further, (I) making the audio anti-counterfeiting information as watermark information includes quantizing the audio information samples into one-dimensional matrix information and then increasing the dimension into two-dimensional matrix information, making the two-dimensional matrix into a Fourier holographic watermark, and performing wavelet transformation. Get the watermark embedding information.

Further, after quantizing the audio information sampling into one-dimensional matrix information, the dimension is increased to two-dimensional matrix information, that is, sampling the monophonic audio information, extracting the symbol information of the audio information, quantizing it into one-dimensional matrix audio symbol information, and converting the one-dimensional matrix Audio symbol information is mapped to a simple discrete grayscale distribution and upscaled to a two-dimensional matrix.

Further, the simple discrete grayscale range of the audio symbol information mapping of the one-dimensional matrix is [0, 0.5, 1], and the dimension increase is performed in the manner of row or column arrangement.

Further, the two-dimensional matrix is made into a Fourier holographic watermark image, that is, the two-dimensional matrix is spread to obtain a spread-spectrum watermark image and multiplied by a random phase to reduce the dynamic range of the hologram, and then Fourier transform is performed on it. The Fourier spectrum, amplitude information and phase information are obtained, and the hologram is obtained according to the optical off-axis hologram transmittance function. Finally, the Fourier holographic watermark is obtained by normalizing the hologram.

Further, the watermark hologram containing the audio symbol information to be produced after the watermark embedding information obtained by wavelet transformation is subjected to Haar wavelet decomposition transformation, and the diagonal high-frequency component information is taken as the watermark information to be embedded.

Further, (II) confirming the embedding position of the watermark in the host image includes constructing the host image into a quaternion imaginary part with a real part of 0 through wavelet decomposition, and performing a quaternion Fourier transform, taking the intermediate frequency of the real part as the watermark. Embed location.

Further, the host image is constructed into a quaternion imaginary part with a real part of 0 through wavelet decomposition, that is, the host image in RGB mode is divided into channels, and each color channel image is subjected to Haar wavelet decomposition, and its low-frequency approximation is taken. The graph information is constructed as the three imaginary parts of a quaternion whose real part is 0.

Further, the performed quaternion Fourier transform is a left quaternion Fourier transform.

Further, (III) transform to obtain a color image containing audio watermark information, that is to add the audio embedding information to the embedding position of the host image, and combine the new real part and intermediate frequency part with the imaginary part to perform inverse quaternion Fourier transform and sum. The anti-counterfeiting image containing audio watermark information is obtained by inverse wavelet transform.

Further, (B) performing information transformation after correcting the image, that is, performing Haar wavelet transform and quaternion Fourier transform on the obtained digital image after correcting the color and size.

Further, the Haar wavelet transform and the quaternion Fourier transform are performed, that is, the three channel images of the image to be detected are respectively subjected to Haar wavelet decomposition, and three low-frequency approximation subgraphs are taken as the three imaginary parts of the quaternion matrix to construct four Arion matrix, quaternion Fourier left transform is performed, and the embedded information of the real part intermediate frequency position is extracted.

Further, (C) restoring the audio anti-counterfeiting information after holographic reproduction, that is, performing holographic reproduction on the extracted embedded information to obtain two-dimensional watermark information, reducing the dimension of the extracted information and selecting a threshold range, and decomposing the data within the threshold range and transforming it. , take the value with the highest positive code rate as the optimal threshold, restore it to symbol information, and perform audio output after post-processing.

Further, the holographic reproduction obtains the two-dimensional watermark information, that is, the extracted embedded information is subjected to inverse Fourier transform, and the reproduced image containing the zero-order image, the original image and the conjugate image can be obtained, and the two-dimensional watermark information is obtained by taking the position information of the original image. .

Further, after dimensionality reduction of the extracted information, the threshold range is selected, that is, the dimension is reduced to one dimension by row or column dimensionality reduction. According to the attenuation trend of the gray histogram, the region with the most obvious attenuation is selected as the primary threshold range.

Further, the data in the threshold range is decomposed and transformed, and the value with the highest positive code rate is taken as the optimal threshold, that is, the data in the threshold range is decomposed n times, and each decomposed data is used as the primary threshold to extract information. Discrete transformation, and compared with the original embedded watermark symbol information, the optimal threshold with the highest positive code rate is obtained according to the positive code rate of logical multiplication.

Further, it is restored to symbol information, and audio output is performed after post-processing, that is, according to the optimal threshold, the extracted one-dimensional information map is restored to symbol information, and audio output is performed after post-processing such as denoising.

The beneficial effects of the present invention are:

1. In the present invention, the original watermark information to be encrypted can be produced by Fourier holographic watermarking technology, which can realize the advantages of encryption and enhanced robustness.

2. The present invention performs multi-resolution analysis on the image through wavelet transform, and embeds the watermark information into the high-frequency components that are insensitive to the human eye extracted by wavelet transform, which is beneficial to the realization of the invisibility of the watermark anti-counterfeiting method, and uses the audio anti-counterfeiting signal as a holographic signal. Watermarks are hidden in printed images, making forgery more difficult.

3. In the present invention, the color image is represented by a quaternion matrix, and the three channel components are comprehensively considered to make it a whole processing method, which is conducive to maintaining the inherent relationship between different channels.

4. The present invention makes full use of the auditory characteristics of the human ear, extracts the audio information hidden in the printed images with the help of mobile terminals such as smart phones, and identifies the authenticity of commodity packaging. Experiential interest in pseudo-discriminatory behavior.

Description of drawings

Fig. 1 is the production process of the audio watermark information hologram of the present invention;

Fig. 2 is the holographic watermarking algorithm embedding flow process of the present invention taking audio as watermark information;

FIG. 3 is the extraction process of the holographic watermarking algorithm using audio as watermark information according to the present invention.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings.

Example 1

As shown in Figures 1-3, a method for hiding and extracting audio anti-counterfeiting signals in a printed image includes two processes of hiding the audio anti-counterfeiting signal in the printed image and extracting the audio anti-counterfeiting information from the printed image; wherein the The audio anti-counterfeiting signal is hidden in the printed image, that is, the audio is used as the watermark anti-counterfeiting information, and the image is used as the host information, including: (I) making the audio anti-counterfeiting information as the watermark information, (II) confirming the watermark embedding position in the host image, (III) ) transform to obtain a color image containing audio watermark information, and use it for printing plate-making output; extracting audio anti-counterfeiting information from the printed image is the hidden inverse process, including: (A) using an imaging device to obtain the image, (B) After correcting Information transformation, (C) restore audio anti-counterfeiting information after holographic reproduction.

Further, (I) making the audio anti-counterfeiting information as watermark information includes quantizing the audio information samples into one-dimensional matrix information and then increasing the dimension into two-dimensional matrix information, making the two-dimensional matrix into a Fourier holographic watermark, and performing wavelet transformation. Get the watermark embedding information.

Further, after quantizing the audio information sampling into one-dimensional matrix information, the dimension is increased to two-dimensional matrix information, that is, sampling the monophonic audio information, extracting the symbol information of the audio information, quantizing it into one-dimensional matrix audio symbol information, and converting the one-dimensional matrix Audio symbol information is mapped to a simple discrete grayscale distribution and upscaled to a two-dimensional matrix.

，归一化后

，选取关键信息处的M1∙M2个长度采样数据的音频信息I，数据不足可用使用0补足，数据超出只截取关键部分信息。Sampling the audio information to obtain audio information A with sampling data L and sampling frequency Fs,

, after normalization

, select the audio information I of the M1∙M2 length sample data at the key information, and use 0 to make up for the lack of data, and only intercept the key part of the information if the data exceeds.

The simple discrete grayscale range of one-dimensional matrix audio symbol information mapping is [0, 0.5, 1], which realizes the transmission of audio information on paper substrates:

The ascending dimension is performed in a row-arranged manner to obtain V(x,y) :

Further, the two-dimensional matrix is made into a Fourier holographic watermark image, that is, the spread-spectrum watermark image f(x, y) obtained by spreading the two-dimensional matrix V(x, y) is multiplied by a random phase φ(x , y) y) to reduce the dynamic range of the hologram, and then Fourier transform it to obtain the Fourier spectrum F(x, y) , the amplitude information A(x, y) and the phase information ϕ(x, y):

其中real[]和imag[]分别为取复波面的实部与虚部操作，根据光学离轴全息图透过率函数得到全息图，全息图透过率函数表达式：

Among them, real[] and imag[] are the operations of taking the real part and imaginary part of the complex wave surface, respectively. The hologram is obtained according to the optical off-axis hologram transmittance function. The hologram transmittance function expression is:

在再现图像时不起作用，且增加了全息图的带宽，用I分量来代替式中的

，推出以下全息函数：because

It does not work when reproducing the image, and increases the bandwidth of the hologram, replace the I component in the formula

, the following holographic function is derived:

where I is a constant, and the normalization operation is such that but:

Finally, the hologram is normalized to obtain the Fourier holographic watermark H(x, y) .

Further, the watermark hologram H(x, y) containing the audio symbol information to be produced after the watermark embedding information is obtained through wavelet transformation is subjected to Haar wavelet decomposition transformation, and the diagonal high-frequency component information w(x, y) is taken as the watermark to be embedded. information.

Further, (II) confirming the embedding position of the watermark in the host image includes constructing the host image into a quaternion imaginary part with a real part of 0 through wavelet decomposition, and performing a quaternion Fourier transform, taking the intermediate frequency of the real part as the watermark. Embed location.

Further, the host image is constructed into a quaternion imaginary part with a real part of 0 through wavelet decomposition, that is, the host image in RGB mode is divided into channels, and each color channel image is subjected to Haar wavelet decomposition, and its low-frequency approximation is taken. The graph information is constructed as the three imaginary parts of the quaternion whose real part is 0. The mathematical formula of the quaternion construction is expressed as: q=a+bi+cj+dk, where a, b, c, d are real numbers, i , j, is an imaginary number, which is a three-dimensional space unit vector perpendicular to each other, and a quaternion is used to represent the original color image, that is, the color image is expressed as:

.

Further, the performed quaternion Fourier transform is the left quaternion Fourier transform, and the left transform formula is:

，其中u=0，1，...，M-1；v=0，1，…，N-1，将得到的实部中频部分作为水印的嵌入位置I(m, n)。

, where u=0, 1, ..., M-1; v=0, 1, ..., N-1, and the obtained real intermediate frequency part is taken as the embedding position I(m, n) of the watermark.

，将新的实部中频部分与虚部结合进行逆四元数傅里叶变换和逆小波变换得到含有音频水印信息的防伪图像，其中逆四元数傅里叶变换公式为：Further, (III) transforms to obtain a color image containing audio watermark information, that is, the audio embedding information is additively embedded into the embedding position of the host image:

, the new real part and the intermediate frequency part are combined with the imaginary part to perform inverse quaternion Fourier transform and inverse wavelet transform to obtain an anti-counterfeiting image containing audio watermark information, where the inverse quaternion Fourier transform formula is:

，其中m=0，1，...，M-1；n=0，1，…，N-1。

, where m=0,1,...,M-1; n=0,1,...,N-1.

Further, (A) obtaining an image with an imaging device, that is, obtaining a digital image on a printed package through a mobile phone.

Further, (B) performing information transformation after correction, that is, performing Haar wavelet transform and quaternion Fourier transform on the obtained digital image after correction of color and size, etc. respectively.

Further, the Haar wavelet transform and the quaternion Fourier transform are performed, that is, the three channel images of the image to be detected are respectively subjected to Haar wavelet decomposition, and three low-frequency approximation subgraphs are taken as the three imaginary parts of the quaternion matrix to construct four Arion matrix, quaternion left Fourier transform:

，其中u=0，1，...，M-1；v=0，1，…，N-1，提取实部中频位置嵌入信息。

, where u=0, 1, ..., M-1; v=0, 1, ..., N-1, extract the real part IF position embedding information.

Further, (C) restoring the audio anti-counterfeiting information after holographic reproduction, that is, performing holographic reproduction on the extracted embedded information to obtain two-dimensional watermark information, reducing the dimension of the extracted information and selecting a threshold range, and decomposing the data within the threshold range and transforming it. , take the value with the highest positive code rate as the optimal threshold, restore it to symbol information, and perform audio output after post-processing.

Further, the holographic reproduction obtains the two-dimensional watermark information, that is, the extracted embedded information is subjected to inverse Fourier transform, and the reproduced image containing the zero-order image, the original image and the conjugate image can be obtained, and the two-dimensional watermark information is obtained by taking the position information of the original image. .

Further, after dimensionality reduction of the extracted information, the threshold range is selected, that is, the dimension is reduced to one dimension by row dimensionality reduction, and the region with the most obvious attenuation is selected as the primary threshold range according to the attenuation trend of the gray histogram.

Further, the data in the threshold range is decomposed and transformed, and the value with the highest positive code rate is taken as the optimal threshold, that is, the data in the threshold range is decomposed n times (more than 2 times), and each decomposed data is used as the primary selection threshold. Simple discrete transformation is performed on the extracted information, and compared with the original embedded watermark symbol information, the optimal threshold with the highest positive code rate is obtained according to the positive code rate of logical multiplication:

e(i) is the same number of logical multiplications in the extracted binary audio information, and is calculated according to:

LP(Logical positive rate )(e)=e(MN)/MN, the logical multiplication positive code rate for extracting information is obtained.

Further, it is restored to symbol information, and audio output is performed after post-processing, that is, according to the optimal threshold, the extracted one-dimensional information map is restored to symbol information, and audio output is performed after post-processing such as denoising:

b and c are the threshold range, and n is the number of segments.

Embodiment 2

The basic principle of this embodiment is the same as that of the first embodiment, but the details are different. A method for hiding and extracting audio anti-counterfeiting signals in a printed image includes hiding the audio anti-counterfeiting signal in the printed image and extracting the audio anti-counterfeiting information from the printed image. There are two processes of extracting from the image; in which the audio anti-counterfeiting signal is hidden in the printed image, that is, the audio is used as the watermark anti-counterfeiting information, and the image is used as the host information, including: (I) Making the audio anti-counterfeiting information as watermark information, (II) Confirming the host The watermark embedding position in the image, (III) transform to obtain a color image containing audio watermark information, and use it for printing plate-making output; extracting audio anti-counterfeiting information from the printed image is the hidden inverse process, including: (A) Using imaging equipment Image acquisition, (B) information transformation after correction, (C) restoration of audio anti-counterfeiting information after holographic reproduction.

Further, (I) making the audio anti-counterfeiting information as watermark information includes quantizing the audio information samples into one-dimensional matrix information and then increasing the dimension into two-dimensional matrix information, making the two-dimensional matrix into a Fourier holographic watermark, and performing wavelet transformation. Get the watermark embedding information.

Further, after quantizing the audio information sampling into one-dimensional matrix information, the dimension is increased to two-dimensional matrix information, that is, sampling the monophonic audio information, extracting the symbol information of the audio information, quantizing it into one-dimensional matrix audio symbol information, and converting the one-dimensional matrix Audio symbol information is mapped to a simple discrete grayscale distribution and upscaled to a two-dimensional matrix.

，选取关键信息处的M1∙M2个长度采样数据的音频信息I，数据不足可用使用0补足，数据超出只截取关键部分信息。Sampling the audio information to obtain audio information A with sampling data L and sampling frequency Fs, after normalization

, select the audio information I of the M1∙M2 length sample data at the key information, and use 0 to make up for the lack of data, and only intercept the key part of the information if the data exceeds.

The simple discrete grayscale range of one-dimensional matrix audio symbol information mapping is [0, 0.5, 1], which realizes the transmission of audio information on paper substrates:

The ascending dimension is performed in a columnar arrangement to obtain V(x,y) :

Further, the two-dimensional matrix is made into a Fourier holographic watermark image, that is, the spread-spectrum watermark image f(x, y) obtained by spreading the two-dimensional matrix V(x, y) is multiplied by a random phase φ(x , y) y) to reduce the dynamic range of the hologram, and then Fourier transform it to obtain the Fourier spectrum F(x, y) , the amplitude information A(x, y) and the phase information ϕ(x, y):

Among them, real[] and imag[] are the operations of taking the real part and imaginary part of the complex wave surface, respectively. The hologram is obtained according to the optical off-axis hologram transmittance function. The hologram transmittance function expression is:

在再现图像时不起作用，且增加了全息图的带宽，用I分量来代替式中的

，推出以下全息函数：because

It does not work when reproducing the image, and increases the bandwidth of the hologram, replace the I component in the formula

, the following holographic function is derived:

=1，R ₀=1，则：where I is a constant, and the normalization operation is such that

=1, R ₀ =1, then:

；

;

Finally, the hologram is normalized to obtain the Fourier holographic watermark H(x, y) .

Further, the watermark hologram H(x, y) containing the audio symbol information to be produced after the watermark embedding information is obtained through wavelet transformation is subjected to Haar wavelet decomposition transformation, and the diagonal high-frequency component information w(x, y) is taken as the watermark to be embedded. information.

Further, (II) confirming the embedding position of the watermark in the host image includes constructing the host image into a quaternion imaginary part with a real part of 0 through wavelet decomposition, and performing a quaternion Fourier transform, taking the intermediate frequency of the real part as the watermark. Embed location.

Further, the host image is constructed into a quaternion imaginary part with a real part of 0 through wavelet decomposition, that is, the host image in RGB mode is divided into channels, and each color channel image is subjected to Haar wavelet decomposition, and its low-frequency approximation is taken. The graph information is constructed as the three imaginary parts of the quaternion whose real part is 0. The mathematical formula of the quaternion construction is expressed as: q=a+bi+cj+dk, where a, b, c, d are real numbers, i , j, is an imaginary number, which is a three-dimensional space unit vector perpendicular to each other, and a quaternion is used to represent the original color image, that is, the color image is expressed as:

。

.

Further, the performed quaternion Fourier transform is the left quaternion Fourier transform, and the left transform formula is:

，其中u=0，1，...，M-1；v=0，1，…，N-1，将得到的实部中频部分作为水印的嵌入位置I(m, n)。

, where u=0, 1, ..., M-1; v=0, 1, ..., N-1, and the obtained real intermediate frequency part is taken as the embedding position I(m, n) of the watermark.

，将新的实部中频部分与虚部结合进行逆四元数傅里叶变换和逆小波变换得到含有音频水印信息的防伪图像，其中逆四元数傅里叶变换公式为：逆变换为：Further, (III) transform to obtain a color image containing audio watermark information, that is, the audio embedding information is additively embedded into the embedding position of the host image:

, the new real part and the intermediate frequency part are combined with the imaginary part to perform inverse quaternion Fourier transform and inverse wavelet transform to obtain an anti-counterfeiting image containing audio watermark information, where the inverse quaternion Fourier transform formula is: The inverse transform is:

，其中m=0，1，...，M-1；n=0，1，…，N-1。

, where m=0,1,...,M-1; n=0,1,...,N-1.

Further, (A) obtaining an image with an imaging device, that is, obtaining a digital image on a printed package by a scanner.

Further, (B) performing information transformation after correction, that is, performing Haar wavelet transform and quaternion Fourier transform on the obtained digital image after correction of color and size, etc. respectively.

Further, the Haar wavelet transform and the quaternion Fourier transform are performed, that is, the three channel images of the image to be detected are respectively subjected to Haar wavelet decomposition, and three low-frequency approximation subgraphs are taken as the three imaginary parts of the quaternion matrix to construct four Arion matrix, quaternion left Fourier transform:

，其中u=0，1，...，M-1；v=0，1，…，N-1，提取实部中频位置嵌入信息。

, where u=0, 1, ..., M-1; v=0, 1, ..., N-1, extract the real part IF position embedding information.

Further, (C) restoring the audio anti-counterfeiting information after holographic reproduction, that is, performing holographic reproduction on the extracted embedded information to obtain two-dimensional watermark information, reducing the dimension of the extracted information and selecting a threshold range, and decomposing the data within the threshold range and transforming it. , take the value with the highest positive code rate as the optimal threshold, restore it to symbol information, and perform audio output after post-processing.

Further, the holographic reproduction obtains the two-dimensional watermark information, that is, the extracted embedded information is subjected to inverse Fourier transform, and the reproduced image containing the zero-order image, the original image and the conjugate image can be obtained, and the two-dimensional watermark information is obtained by taking the position information of the original image. .

Further, after dimensionality reduction of the extracted information, the threshold range is selected, that is, the dimension is reduced to one dimension according to the column dimensionality reduction method, and the region with the most obvious attenuation is selected as the primary threshold range according to the attenuation trend of the gray histogram.

Further, the data in the threshold range is decomposed and transformed, and the value with the highest positive code rate is taken as the optimal threshold, that is, the data in the threshold range is decomposed n times (more than 3 times), and each decomposed data is used as the primary selection threshold. Simple discrete transformation is performed on the extracted information, and compared with the original embedded watermark symbol information, the optimal threshold with the highest positive code rate is obtained according to the positive code rate of logical multiplication:

e(i) is the same number of logical multiplications in the extracted binary audio information, and is calculated according to:

LP(Logical positive rate )(e)=e(MN)/MN, the logical multiplication positive code rate for extracting information is obtained.

Further, it is restored to symbol information, and audio output is performed after post-processing, that is, according to the optimal threshold, the extracted one-dimensional information map is restored to symbol information, and audio output is performed after post-processing such as denoising:

b and c are the threshold range, and n is the number of segments.

Thus, the present invention also has the following beneficial effects:

1. In the present invention, the original watermark information to be encrypted can be produced by Fourier holographic watermarking technology, which can realize the advantages of encryption and enhanced robustness.

2. The present invention performs multi-resolution analysis on the image through wavelet transform, and embeds the watermark information into the high-frequency components that are insensitive to the human eye extracted by wavelet transform, which is beneficial to the realization of the invisibility of the watermark anti-counterfeiting method, and uses the audio anti-counterfeiting signal as a holographic signal. Watermarks are hidden in printed images, making forgery more difficult.

3. In the present invention, the color image is represented by a quaternion matrix, and the three channel components are comprehensively considered to make it a whole processing method, which is conducive to maintaining the inherent relationship between different channels.

4. The present invention makes full use of the auditory characteristics of the human ear, extracts the audio information hidden in the printed images with the help of mobile terminals such as smart phones, and identifies the authenticity of commodity packaging. Experiential interest in pseudo-discriminatory behavior.

Claims (8)

1. a method for concealing and extracting audio anti-counterfeiting signal in printed image, comprising audio anti-counterfeiting signal hidden in printed image and two processes of audio anti-counterfeiting information being extracted from printed image, it is characterized in that: audio anti-counterfeiting signal is hidden In the printed image, it refers to the audio as the watermark anti-counterfeiting information and the image as the host information, including: (I) making the audio anti-counterfeiting information as the watermark information, (II) confirming the watermark embedding position in the host image, (III) transforming to obtain Color images containing audio watermark information, and used for printing plate-making output; extracting audio anti-counterfeiting information from printed images refers to the inverse process of hiding audio anti-counterfeiting signals in printed images, including: (A) using imaging equipment to obtain images, ( B) Perform information transformation after image correction, (C) restore audio anti-counterfeiting information after holographic reproduction; The (I) making the audio anti-counterfeiting information as watermark information includes quantizing the audio information samples into one-dimensional matrix information and then increasing the dimension into two-dimensional matrix information; watermark embedded information; The audio information sampling is quantized into one-dimensional matrix information and then upscaled to two-dimensional matrix information, that is, the monophonic audio information is sampled, the symbol information of the audio information is extracted, quantized into one-dimensional matrix audio symbol information, and the one-dimensional matrix audio The symbolic information is mapped to a simple discrete grayscale distribution and raised to a two-dimensional matrix; The simple discrete grayscale range of the one-dimensional matrix audio symbol information mapping is [0, 0.5, 1], and realizes the transmission of audio information on the paper substrate:

The ascending dimension is arranged in rows or columns to obtain V(x,y) :

; The two-dimensional matrix is made into a Fourier holographic watermark image, that is, after the two-dimensional matrix is spread, a spread-spectrum watermark image is obtained and multiplied by a random phase to reduce the dynamic range of the hologram, and then Fourier transform is performed on it to obtain Fourier spectrum, amplitude information and phase information, obtain the hologram according to the optical off-axis hologram transmittance function, and finally normalize the hologram to obtain the Fourier holographic watermark; The watermark hologram containing the audio symbol information to be produced after the wavelet transform obtains the watermark embedding information is subjected to Haar wavelet decomposition transformation, and the diagonal high-frequency component information is taken as the watermark information to be embedded. 2. The method for hiding and extracting audio anti-counterfeiting signals in a printed image as claimed in claim 1, wherein (II) confirming the watermark embedding position in the host image comprises constructing the host image into a real part through wavelet decomposition The imaginary part of the quaternion is 0, and the quaternion Fourier transform is performed, and the intermediate frequency of the real part is taken as the watermark embedding position. 3. the method for concealing and extracting audio anti-counterfeiting signal in printed image as claimed in claim 2, it is characterized in that: described host image is constructed by wavelet decomposition to be the quaternion imaginary part whose real part is 0, namely under RGB mode The host image is divided into channels, and each color channel image is subjected to Haar wavelet decomposition, and its low-frequency approximation sub-image information is constructed as the three imaginary parts of the quaternion whose real part is 0; The performed quaternion Fourier transform is a left quaternion Fourier transform. 4. The method for hiding and extracting audio anti-counterfeiting signals in a printed image as claimed in claim 1, characterized in that: said (III) transforming to obtain a color image containing audio watermark information is to add the audio embedding information to the host image The new real part and the intermediate frequency part are combined with the imaginary part to perform inverse quaternion Fourier transform and inverse wavelet transform to obtain an anti-counterfeiting image containing audio watermark information. 5 . The method for hiding and extracting audio anti-counterfeiting signals from a printed image as claimed in claim 1 , wherein (B) performing information transformation after image correction, that is, after color and size correction of the obtained digital image. 6 . Haar wavelet transform and quaternion Fourier transform are carried out respectively. 6. the method for hiding and extracting audio anti-counterfeiting signal in printed image as claimed in claim 5, it is characterized in that: described carrying out Haar wavelet transform and quaternion Fourier transform namely three channel images of the image to be detected are respectively Perform Haar wavelet decomposition, take three low-frequency approximation subgraphs as the three imaginary parts of the quaternion matrix, construct the quaternion matrix, perform the left Fourier transform of the quaternion, and extract the intermediate frequency position embedding information of the real part. 7 . The method for hiding and extracting audio anti-counterfeiting signals in a printed image according to claim 1 , wherein: (C) restoring the audio anti-counterfeiting information after holographic reproduction is obtained by holographic reproduction of the extracted embedded information. 8 . For two-dimensional watermark information, the extracted information is dimensionally reduced and then the threshold range is selected, the data within the threshold range is decomposed and transformed, and the value with the highest positive code rate is taken as the optimal threshold value, which is restored to symbol information, and the audio is processed after post-processing. output. 8. The method for hiding and extracting audio anti-counterfeiting signals in a printed image as claimed in claim 7, wherein the holographic reproduction obtains the two-dimensional watermark information, that is, the extracted embedded information does inverse Fourier transform, and can obtain The reconstructed image containing the zero-order image, the original image and the conjugated image, obtain the two-dimensional watermark information by taking the position information of the original image; After the extracted information is reduced in dimension, the threshold range is selected, that is, the dimension is reduced to one dimension by row or column dimension reduction, and the region with the most obvious attenuation is selected as the primary selection threshold range according to the attenuation trend of the gray histogram; The data in the threshold range is decomposed and transformed, and the value with the highest positive code rate is taken as the optimal threshold, that is, the data in the threshold range is decomposed n times, and each decomposed data is used as the primary selection threshold to perform simple discrete extraction of information. Transform and compare with the original embedded watermark symbol information, and obtain the optimal threshold with the highest positive code rate according to the positive code rate of logical multiplication.

Images